ResearchBenchmarksRetrieval Augmented Generation on UltraDomain multi-hop n=77Follow36.8F1 ScoreContextRAG34.9635.8836.837.72May 19, 2026Evaluation ResultsMethodMethodLinksF1 ScoreContextRAGTasks=77 (multi-hop su...Tasks=77 (multi-hop subset)2026.0536.8